Workshop - Legal and Ethical Challenges for Driverless Cars and Smart Roads (October 2017)
- Slide Presentation at Bath Digital Festival
- Top 6 Legal and Ethical Challenges
- Cultural Differences
- Further information
Slide Presentation at Bath Digital Festival
October 2017, Bath Digital Festival (United Kingdom)
Read more »
- Driverless cars
- Level 4 and Level 5 on the autonomous vehicle scale - no driver at all, with a computer making all decisions once a navigation target and some basic rules are set
- Smart roads
- Road networks that interact with all vehicles and road infrastructure (traffic lights, speed limits etc.)
- Most vehicles from 2018 will be fitted with an automatic transponder and older vehicles can be retro-fitted
- Modern roads have some smart infrastructure built in - more on the way
- Driverless cars and smart roads are a rapidly emerging technology
- The law is famously poor at keeping up with new technology
Top 6 Legal and Ethical Challenges
- 1. Privacy
- 2. Safety and Selection
- 3. Liability
- 4. Cultural Differences
- 5. Traffic Priorities
- 6. Trust
- Constant broadcast
- A typical transponder broadcasts vehicle type and make, speed, braking, acceleration, indicators etc.
- The transponder is identified by a signed digital certificate (to avoid fraud and impersonation)
- Access requests
- Requests will be common for both live and historical movement and location data - from law enforcement, lawyers in civil disputes, general traffic management etc.
- Privacy law
- Privacy law currently relies on notice and consent (difficult for vehicles) and in any case provides huge exceptions for law enforcement, emergencies etc.
- Privacy by design
- A Privacy Impact Assessment can often identify solutions - for example, in one ‘privacy friendly’ option a bundle of certificates for each transponder can be randomly shuffled
- Safety and selection
- Often called ‘The Trolley Problem’ - where an accident resulting in a fatality is unavoidable, how do you decide who should die?
- You can even ‘play’ a game based on the Trolley Problem at MIT:
- Case study: Germany (Guidelines 2017)
- Self-driving cars must prioritise human life over property and animals.
- Self-driving cars must do the least amount of harm if put into a situation where hitting a human is unavoidable
- Self-driving cars must not discriminate based on age, gender, race, disability, or any other observable factors.
- A common approach to determining liability is to assess which party has the greatest ability to avoid damage. A supplementary test is which party has the greatest ability to compensate for any damage
- Vehicle manufacturers
- In order to build confidence in driverless cars, some manufacturers have offered an indemnity for any damage (but check the fine print!)
- Manufacturers are unlikely to be able to avoid liability for any damage resulting from their negligence in design or implementation (due to consumer protection laws - but these laws differ from country to country)
- Owners may be pressured to accept some liability (e.g. in contracts). There are protections against unfair contract terms in the UK.
- Compulsory insurance is likely to be the long term solution to liability issues for driverless cars - the manufacturer indemnities are more akin to an introductory offer or stunt
4. Cultural Differences
- Driving is cultural
- More than just the left / right divide - there are numerous national and regional differences in driving behavior and traffic management
- Most approaches are based on customs or etiquette, but some are enshrined in law
- Managing cultural differences
- This will be challenging if algorithms for driverless cars are developed in just a handful of jurisdictions, or if AI is based on data initially obtained from just one culture.
- Recent developments in AI have demonstrated an ability for AI to develop new knowledge itself, raising questions about ‘who is really in charge’.
- The most difficult phase will be when driverless cars and traditional cars have to share limited road space
- Smart roads will have the ability to prioritise specific vehicles and manage overall traffic patterns
- General recognition that emergency vehicles will receive priority from driverless cars and smart road infrastructure (e.g, traffic signals)
- Some successful pilots of smart roads allowing individual heavy goods vehicles (and convoys) to reach a destination with minimal stopping (reduces environmental impact and road maintenance)
- Who determines these priorities?
- Significant potential for conflict, bias and influence
- Potential for entrenching privilege / disadvantage
- AI and priorities
- In one AI experiment the AI becomes more aggressive as the challenge becomes more competitive
- Trust is a key issue when you are relying on an algorithm to make key decisions
- Users may not be able to see or understand the details of the algorithm
- The algorithm may make selections or priorities without the consumer being aware
- Trust in the vehicle / transport sector is in crisis
- Many industry players are ‘disrupters’ who have gone to great lengths to avoid or undermine regulation
- For example, Uber is a leading player in the driverless car sector. They have been the subject of a series of controversies, including:
- Two major privacy breaches where senior management directed staff to place ‘opponents’ (journalists) under surveillance and even directly threatened journalists with revealing their personal data
- Revelations that Uber had developed and used specific software (Greyball) to identify and avoid regulatory staff (e.g. inspectors)
- Numerous vehicle manufacturers have also been caught up in the emissions testing fraud scandal
- Overall, driverless cars and smart roads have the potential to deliver significant benefits
- Driverless cars are not affected by fatigue, alcohol, health conditions and distractions
- Smart roads and transponders allow vehicles to ‘see’ traffic hidden by hills, corners, fog, snow and blinding light
- Traffic management and vehicle priority (e.g. emergency vehicles) are enhanced by smart roads and transponders
- However, key issues will require careful management, including direct intervention and regulation
- AI and algorithms need to be transparent and subject to rules and restraints
- The core approach should be:
- “Even if there is no driver behind the wheel, a human is always in charge”
- US database of driverless car legislation
- Privacy Impact Assessment on Smart Roads (Australia, 2017)
- Germany: Guidelines on Driverless Cars and Ethics (2017)
October 2017, Bath Digital Festival (United Kingdom)
Read more »